A variety of systems are used for automated inspection of semiconductor wafers, in order to detect defects, particles and/or patterns on the wafer surface as part of a quality assurance process in semiconductor manufacturing processes. It is a goal of current inspection systems to have high resolution and high contrast imaging in order to provide the reliability and accuracy demanded in sub-micron semiconductor manufacturing processes. However, it is also important to have a high-speed process that permits a large volume throughput so that the quality and assurance processes do not become a bottleneck in the wafer production process. Accordingly, the optical inspection systems must use shorter wave lengths, higher numerical aperture optics and high density image capture technology in order to enable the processing of data from such systems at sufficiently high rates that will satisfy the desired product throughput requirements.
A conventional imaging architecture that is used in wafer inspection systems at this time utilizes multi spot scanning laser for high-speed imaging. However, the data rates achievable by such architectures are limited by the physical constraints that arise due to limitations in the speed and quality of the single laser beam, the applicable optical system and related detection devices. For example, the single laser acting as a point light source is focused as a spot onto the object under inspection and is scanned across the surface of the object, which may be stationary or moved on a stage mechanism in coordination with the scan. The reflected light from the object is then imaged onto a detector, which generates pixel data from the scanning process.
The detector may be a photo multiplier detector (PMT) or a CCD array, whose individual elements are positioned to receive the reflected light as the beam is scanned and be read our serially, in a conventional fashion. While a high resolution may be obtained from such point source illumination, the requirement to scan each point in the field in order to construct a viewable image subjects the system to a limitation on its throughput.
The detector has to image the entire scan path of the spot and may collect stray light or other noises. The scanning of the single laser beam may be accomplished by a rotating mirror system, as seen in U.S. Pat. No. 5,065,008 or an acousto-optic cell. However, these single spot scanning architecture necessarily have a limited speed and are possibly subject to scan aberrations, low illumination brightness and potential thermal damage to the object when high brightness laser sources are used. The high data rates required to inspect the submicron structures of current semiconductor products cannot be achieved, even when a stage-type scanning system is used that moves the object relative to a fixed illumination and image location while a synchronized scanning pattern is produced by moving the single point of light over an area at the fixed location.
One way to increase the throughput of the inspection is to scan the object with a rectangular grid of beams, wherein the scan axis is parallel to the columns of the grid.
When the object is scanned with multiple beams the detector may also suffer from cross talk between the multiple beams.
Accordingly, there is a need for an object scanning system that will improve object throughput, while maintaining or even improving the reliability and accuracy of the data collected during the scan of an object, whether in a stationary or stage-type system.